AI fan wikis riddled with errors are muddying the waters of games culture
AI-generated fan wikis and guide sites are flooding game searches with plausible but error-ridden information, undermining long‑standing community efforts to document game mechanics accurately. Commenters argue this is part of a wider decline in online information quality as large language models make it trivial to mass‑produce convincing text, and they debate possible responses ranging from human‑only or heavily moderated sites to new identity and trust systems for the web. Many are pessimistic about preserving an open, low-friction internet that remains both usable and reliably human-authored.
Perceived AI Degradation of Online Information
- Many see AI-generated “slop” as accelerating an existing decline in online quality: discourse, documentation, and reference material are all cited.
- Concern that once AI is trained on AI output, information ecosystems will undergo “bit rot,” with recursive degradation of quality.
- Some argue the open web will go the way of spam-ridden email/Usenet, becoming saturated with low-value automated content.
Game Wikis and Fan Communities
- AI “wikis” are described as thin, plausibly written but error-filled sites that confuse players and pollute search results.
- Legacy commercial wiki platforms are criticized as already low-quality, so AI slop is seen as piling onto an existing problem.
- Some argue this damages fan cultures and communities built around careful, human curation; others joke that players may now need to “actually play the game.”
Quality of Code and Technical Content
- One side claims LLMs will improve average code quality and avoid entire bug classes.
- Others report catching obvious security bugs in AI-suggested code and worry about code/comments/docs whose human vs AI origin is unknowable.
- This ambiguity complicates maintenance, since you no longer know if odd decisions reflect real intent or fluent nonsense.
Search Engines and “Slop” Websites
- Multiple anecdotes describe it becoming hard to find reliable web guidance (e.g., for construction work); AI-filled “blogs” now dominate results.
- DuckDuckGo is singled out as particularly polluted; Google is seen as sliding but still slightly better.
- Calls for a search engine that filters out AI slop and possibly paywalls; some mention moving to YouTube, books, or old forum posts instead.
Proposed Responses and Their Limits
- Ideas: human-only or high-friction contribution sites; strict moderation and bans on LLM use (Wikipedia cited positively); invite-only communities; making AI use more expensive.
- Skeptics argue any scalable, non–line-by-line system can be gamed with AI, and incentives will steadily corrupt any “new web.”
- Some foresee a small, curated “if you know, you know” web for serious users, with the mass web dominated by AI output.
Identity, Trust, and Web-of-Trust Ideas
- One camp advocates strong identity (possibly biometric) to prove “human-ness” and fight bot floods, even at the cost of anonymity.
- Others strongly oppose this as “surveillance capitalism,” arguing we should preserve pseudonymity and accept that we’ll sometimes talk to bots.
- Cryptographic signatures and webs of trust are proposed: you trust content if you trust the signer, directly or via social chains.
- Counterpoints note humans can sign AI-written text, keys can be stolen or coerced, and reputation systems are abusable.
LLMs as Powerful Analysis Tools
- Several commenters note frontier LLMs can be excellent at reverse engineering games: disassembling binaries, reading game files, and extracting exact mechanics or loot formulas that wikis and forums get wrong.
- They see a distinction between high-end models doing deep analysis on primary data (good) and cheap models generating surface-level prose from existing web content (bad).
Return to Offline / Pre-AI Sources
- Some users are archiving pre-AI copies of Wikipedia/OSM, likening them to “low-background steel” for future reference.
- Others recommend libraries, paper books, and offline expertise as increasingly important, especially for stable domains like DIY.